void Labeling::read(const cv::Mat3b &src)
{
std::map<cv::Vec3b, short, Vec3bCompare> palette;
palette[cv::Vec3b(0, 0, 0)] = -1; // always have black background label
// find all colors
cv::Mat3b::const_iterator its;
for (its = src.begin(); its != src.end(); ++its)
palette[*its] = -1;
// assign labelColors based on map order (somewhat canonical label indices)
labelColors.clear();
std::map<cv::Vec3b, short, Vec3bCompare>::iterator itp;
for (itp = palette.begin(); itp != palette.end(); ++itp)
{
itp->second = labelColors.size(); // index into labelColors
labelColors.push_back(itp->first);
}
labelcount = labelColors.size();
// assign the color indices to the label matrix
labels = cv::Mat1s(src.rows, src.cols);
cv::Mat1s::iterator itl;
for (its = src.begin(), itl = labels.begin();
its != src.end(); ++its, ++itl) {
*itl = palette[*its];
}
/* Special case: only one white label stored in RGB image (stupid).
We don't want to use white color, it confuses the user. */
if ((labelcount == 2) && (labelColors[1][0] == labelColors[1][1])
&& (labelColors[1][1] == labelColors[1][2]))
labelColors[1] = cv::Vec3b(0, 255, 0);
}
cv::Mat3b OpenCVUtil::pad_image(const cv::Mat3b& image, int paddingSize)
{
cv::Mat3b paddedImage = cv::Mat3b::zeros(image.rows + paddingSize * 2, image.cols + paddingSize * 2);
cv::Rect roi(paddingSize, paddingSize, image.cols, image.rows);
image.copyTo(paddedImage(roi));
return paddedImage;
}
void RGBDImage :: fillRgbFromUserLabels(cv::Mat3b& img) const
{
if (!m_user_labels.data)
return;
const Vec3b colors[] = {
Vec3b(255,0,0),
Vec3b(255,255,0),
Vec3b(255,0,255),
Vec3b(255,255,255),
Vec3b(0,255,0),
Vec3b(0,255,255),
Vec3b(0,0,255),
};
int nb_colors = sizeof(colors) / sizeof(Vec3b);
img.create(m_user_labels.size());
for_all_rc(m_user_labels)
{
int label = m_user_labels(r,c);
if (label == 0)
img(r,c) = Vec3b(0,0,0);
else
img(r,c) = colors[label%nb_colors];
}
}
void HSColorModel :: show(cv::Mat3b& hist_img) const
{
ntk_assert(m_histogram.data, "Model was not build.");
ntk_assert(m_histogram.dims == 2, "This is for 2d histograms");
int scale = 10;
int x_bins = m_histogram.size[0];
int y_bins = m_histogram.size[1];
double min_value = 0, max_value = 0;
minMaxLoc(m_histogram, &min_value, &max_value, 0, 0 );
hist_img.create(x_bins*scale,y_bins*scale);
hist_img = Vec3b(0,0,0);
for(int h = 0; h < x_bins; h++ )
{
for(int s = 0; s < y_bins; s++ )
{
float bin_val = m_histogram(h,s);
int intensity = cvRound(bin_val*255.0/max_value);
rectangle(hist_img,
Point( h*scale, s*scale ),
Point( (h+1)*scale - 1, (s+1)*scale - 1),
CV_RGB(intensity,intensity,intensity),
CV_FILLED );
}
}
}
void tgTextureModel::TextureFromImage(const tgCamera& cam, const cv::Mat3b& image)
{
ComputeFaceNormals();
m_face_tex_id.clear();
std::vector<bool> vertex_used(m_vertices.size(), false);
for (size_t i = 0; i < m_faces.size(); i++)
{
TomGine::tgFace &f = m_faces[i];
for (size_t j = 0; j < f.v.size(); j++)
{
const unsigned &vidx = f.v[j];
if (vertex_used[vidx])
continue;
vertex_used[vidx] = true;
TomGine::tgVertex &v = m_vertices[vidx];
v.texCoord = cam.GetTexCoords(v.pos);
v.texCoord.y = 1.0f - v.texCoord.y; // cv::Mat to OpenGL
}
m_face_tex_id.push_back(0);
}
m_tex_cv.resize(1);
image.copyTo(m_tex_cv[0]);
}
bool
ImageWidget :: setImage(QImage& image, const cv::Mat3b& im)
{
bool geometryUpdated = false;
if (image.isNull()
|| image.width() != im.cols
|| image.height() != im.rows)
{
image = QImage(im.cols, im.rows, QImage::Format_RGB32);
geometryUpdated = true;
}
for (int r = 0; r < im.rows; ++r)
{
QRgb* ptr = (QRgb*) image.scanLine(r);
const uchar* cv_ptr = im.ptr(r);
for (int i = 0; i < im.cols; ++i)
{
int rgb = 0xff << 24;
rgb |= (*cv_ptr++);
rgb |= ((*cv_ptr++) << 8);
rgb |= ((*cv_ptr++) << 16);
*ptr++ = rgb;
}
}
return geometryUpdated;
}
std::vector<cv::Point2f> findLabelPositions(const cv::Mat3b& img, double threshold, bool visualize)
{
std::vector<LabelContour> labels = findLabelContours(img, threshold, visualize);
printf("Num labels: %lu\n", labels.size());
fflush(stdout);
if (visualize)
{
cv::Mat3b canvas = img * 0.25f;
drawLabels(canvas, labels, cv::Scalar(255, 255, 255));
cv::imshow("detected labels", canvas);
}
// Connect labels with each other in order to associate them.
std::vector<std::vector<LabelContour>> grouped_labels;
std::vector<std::vector<cv::Point2f>> spatial_indices;
std::tie(grouped_labels, spatial_indices) = connectLabelContours(labels, visualize);
Camera cam;
try
{
cam = solveCamera(grouped_labels, spatial_indices, img.size());
}
catch (cv::Exception)
{
return {};
}
return projectCube(cam);
}
void apply_mask(cv::Mat3b& im, const cv::Mat1b& mask)
{
if (!mask.data)
return;
ntk_assert(im.size() == mask.size(), "Wrong mask size");
for_all_rc(im)
if (mask(r,c) == 0)
im(r,c) = Vec3b(0,0,0);
}
void BlendScreen(cv::Mat3b& target_image, const cv::Mat1b& binary_image, cv::Vec3b color) {
assert(target_image.size() == binary_image.size());
cv::Size loop_size = target_image.size();
if(target_image.isContinuous() && binary_image.isContinuous()) {
loop_size.width *= loop_size.height;
loop_size.height = 1;
}
for(int i = 0; i < loop_size.height; ++i) {
cv::Vec3b* target = target_image.ptr<cv::Vec3b>(i);
const unsigned char* binary = binary_image.ptr<unsigned char>(i);
for(int j = 0; j < loop_size.width; ++j) {
if(!binary[j]) {
target[j][0] = target[j][0] + color[0] - (target[j][0] * color[0]) / 255;
target[j][1] = target[j][1] + color[1] - (target[j][1] * color[1]) / 255;
target[j][2] = target[j][2] + color[2] - (target[j][2] * color[2]) / 255;
}
}
}
}
void ImageWidget :: setImage(const cv::Mat3b& im)
{
if (m_image.width() != im.cols
|| m_image.height() != im.rows)
m_image = QImage(im.cols, im.rows, QImage::Format_RGB32);
for (int r = 0; r < im.rows; ++r)
{
QRgb* ptr = (QRgb*) m_image.scanLine(r);
const uchar* cv_ptr = im.ptr(r);
for (int i = 0; i < im.cols; ++i)
{
int rgb = 0xff << 24;
rgb |= (*cv_ptr++);
rgb |= ((*cv_ptr++) << 8);
rgb |= ((*cv_ptr++) << 16);
*ptr++ = rgb;
}
}
update();
}
std::vector<cv::Scalar> readLabelColors(const cv::Mat3b img, const std::vector<cv::Point2f>& points)
{
std::vector<cv::Scalar> label_colors;
cv::Rect img_rect(cv::Point(0,0), img.size());
for (const auto& p :points)
{
if (img_rect.contains(p))
{
label_colors.push_back(img(p));
}
}
if (label_colors.size() == 9*3)
{
return label_colors;
}
else
{
return {};
}
}